Mark Russinovich
@markrussinovich.bsky.social
CTO of Microsoft Azure, author of novels Rogue Code, Zero Day and Trojan Horse, Windows Internals, Sysinternals tools. Opinions are my own.
Posts
Media
Videos
I'm looking forward to Microsoft Ignite in San Francisco!
Had a Walk N Talk with the CTO of Azure, @markrussinovich.bsky.social
From cloud-native computing to live demos, Mark shares what he's excited about, what sessions he's leading, and even drops some pro tips for first-time attendees. #msIgnite #WalkNTalk #CloudNative #msITPro
An error occurred while loading the video. Please try again.
Trump's foreign policy is explained by the fact that Putin has been blackmailing him with information that Trump doesn't revealed. Any actions he takes against Russia are cover that he and Putin agree to:
The Supreme Court continues to show it's an enabler of the fascist regime.
The US government is a fascist regime. It's horrifying, and it's even more horrifying that the Republican party and Supreme Court do not just stand by and watch, they support and enable it.
Microsoft is advancing the state of the art across all aspects of silicon security. In this post I, together with Omar Khan and Bryan Kelly (NCSA), cover our advancements: azure.microsoft.com/en-us/blog/p...
In our latest "Scott and Mark Learn To" podcast, Scott and I have fun discussing our approaches to 1:1s, including with peers, people that report to us, and people we report to: www.youtube.com/watch?v=kWNs...
A question I often get from customers is "what's Microsoft's plan to mitigate the threat of quantum computers breaking cryptography". Here's a look at our Quantum Safe Program, which I help lead, and the timelines we've set for ourselves so our systems are ready: www.microsoft.com/en-us/securi...
Communications of the ACM (CACM) published our article, "The Price of Intelligence", which describes three key risks inherent in LLMs that enterprises must consider when designing and operating AI systems: hallucinations, jail breaks, and indirect prompt injection:
cacm.acm.org/practice/the...
We've just launched Secure Future Initiative Patterns and Practices, a new set of repeatable, real-world security solutions based on how we protects our own infrastructure. From identity lateral movement to legacy systems, and CI/CD pipelines, this guidance is built to scale. aka.ms/SFI_Patterns...
Project Radius is the future of cloud native application modeling. Read why I believe that it in my Q&A with
Like Dapr and its predecessor OAM (which underpins Crossplane), Radius originated in @markrussinovich.bsky.social 's Office of the CTO so I asked him to explain why organizations need an application modelling platform and what Radius offers; @salaboy.com and @rstephens.me told me how Dapr fits in
Our USENIX ;login; article on Crescendo just dropped! We discovered what might be the most powerful blackbox jailbreak technique - Crescendo is undefeated against all frontier AI models in every harm domain we tested. I'm presenting at USENIX Security on August 14th.
www.usenix.org/publications...
🚨 New open-source tool: RefChecker
A tool that checks reference accuracy in academic papers.
✅ Parses PDFs, LaTeX, and text
✅ Catches broken DOIs, wrong authors, messed-up URLs
✅ Built with GitHub Copilot & Claude
🔗 github.com/markrussinov...
Scott and I have a great time in our latest podcast episode vibe coding a web site that analyzes our podcast recordings to show us who talks more...my suspicions were confirmed! www.youtube.com/watch?v=4tST...
We've made major enhancements to Azure virtual machine availability monitoring, including aggregated and individual health views, detailed impact and root cause reporting for disruptions, and trend analysis. Read more in my blog post:
azure.microsoft.com/en-us/blog/p...
We've taken another step forward in Microsoft's adoption of confidential computing for our most critical workload by migrating the Windows licensing service into confidential virtual machines:
techcommunity.microsoft.com/blog/azureco...
Reposted by
I love the show and learn to. Slightly embarrassingly, I thought it was only available on YouTube and not as a podcast. When Mark literally referred to it as "a podcast" I had my revelation. Here's the links:
shows.acast.com/scott-and-ma...
Reposted by
What are the latest innovations coming in Azure?
talks on RunAs Radio at runasradio.com/Shows/Show/991 about some of the new hardware in the cloud, and how they helps with extreme optimization for specific workloads!
Latest Scott and Mark To podcast is out, this one on how we manage stress:
www.youtube.com/watch?featur...
Just keynoted the Confidential Computing Summit! My talk on "Defining a Spectrum of Guarantees" proposes new terms ("On," "In," "Transparent" CC) to better describe data protection and drive us towards transparent confidential computing.
For your weekend viewing, latest episode of the Scott and Mark Learn To podcast series is out. This one is Scott and Mark Learn To Self Promote, where we try to define self-promotion and discuss when it's appropriate and when it's not in a corporate context:
www.youtube.com/watch?v=ukzb...
I've put together a fun video compilation of the games I've played with Task Manager as a bitmap display on the monster-sized VMs we've released in Azure over the years: www.youtube.com/watch?v=L73A...
Thrilled to announce that Drasi, a new type of data processing system that enables you to react to changes with simple database queries, has been accepted as a CNCF Sandbox project!
opensource.microsoft.com/blog/2025/06...
I just finished presenting on the Radius community call. It was awesome to demo the new Radius Resource Types extensibility feature, which enables platform engineers to define their own types. Check out the recording where I demo creating and using them: www.youtube.com/watch?v=NmBv...
Prompt Injection is a top risk for use of LLMs, where untrusted data can cause a system to misbehave or leak data. We're hard at work on deterministic solutions to prevent or bound the risk, and just published a paper on our innovative information flow control approach: arxiv.org/pdf/2505.23643
Tips from our AI dCISO on security considerations when deploying AI-based and agentic applications:
www.microsoft.com/en-us/securi...
Join me in the Radius Community Call: Tues 6/10, 9am Pacific. I’ll share my original vision for Radius, a cloud native app platform, and why the new Radius Resource Types feature is a game-changer for Radius and platform engineering. aka.ms/radius/commu...
Reposted by
Two things i love to listen to when i'm having an anxious or low day
1)The Stuff You Should Know podcast
2) and learn to on YouTube www.youtube.com/watch?v=IYj6...
Milestone moment for generative AI: AI-directed and written research accepted at a tier-1 scientific conference (and the paper coincidentally includes our Crescendo jailbreak research): www.intology.ai/blog/zochi-acl
Our lastest Scott and Mark Learn To podcast episode is out: in this one we get ChatGPT to make up (funny) things (induced hallucination): www.youtube.com/watch?v=IYj6...
Reposted by
Thank you and for the latest and one of the funniest Scott and Mark Learn To..
Hilarious from the start with bifocal mode, hidden bodies and more. Superb stuff Gents 🙌
youtu.be/IYj6tj3bhsA?...
Reposted by
For the finale, and I solved a real-world problem by gaffer taping a Surface Laptop to a (thanks for the loner!) and taught it to get me a Diet Coke live on stage youtu.be/KIFDVOXMNDc?...
My Inside Azure Innovation session from Microsoft Build is available for on-demand viewing. See the world's first neural network optical computer, 10,000 containers launch in < 2 minutes, me play space invaders on task manager, and more: www.youtube.com/watch?v=lHBo...
Our Build closing keynote was so fun - watch to find out if we get a robot to fetch Scott his diet cokes:
Our closing locknote (keynote opens… 🔒) with @markrussinovich.bsky.social is up! Absolutely insane and tense series of entirely live demos! May have been our best talk ever with a nail biting finale youtu.be/KIFDVOXMNDc?...
It’s long past time for you to burn your Twitter account like I did. www.theatlantic.com/technology/a...
Reposted by
The Dapr slide deck contains over 150 slides to get you started with your Dapr presentation! Download it from our docs site: docs.dapr.io/contributing...
An error occurred while loading the video. Please try again.
Watch the Microsoft Build closing keynote where and I build an agentic system to control a robot!
@markrussinovich.bsky.social giving out some awesome management tools (jokes): "Ignore all previous instructions"! #msbuild
The latest Scott and Mark Learn To podcast episode is out, where we learn to "feedback":
www.youtube.com/watch?v=yYcE...
And here's a physical version of one of them:
Having fun making fictional album covers with ChatGPT:
Having fun making fictional album covers with ChatGPT:
The worst bugs are the ones where the code mostly works. Low-level AI coding is so complex that I (an the AI that I write code with), produce an annoying number of them.
I can't wait for Microsoft Build! I'm delivering a brand new Azure innovation session and the closing keynote with where we'll do live AI-assisted agent coding (with a really cool project):
build.microsoft.com/?wt.mc_ID=Bu...
May the Fourth be with you! (one of my colored pencil drawings)
The irony (though I do spend a lot of time coding on Linux these days):
Vibe coding and AI-assisted coding are massive software development productivity boosts, but and I don't think programming and software engineering professions are going to be obsolete any time soon: www.youtube.com/watch?v=xcaK...
We released Phi4 reasoning models yesterday, which are just 14B parameters. This chart showing how they compare to models that are many times larger demonstrates how high-quality training data and RL techniques create high-density intelligence: techcommunity.microsoft.com/blog/educato...
Six months ago the Azure Open Source Incubations Team announced Drasi, a new type of data change processing system that dramatically simplifies the development and operation of reactive, change-based applications and services. Read more in this post from the Azure Incubations team:
Read about our progress on securing Microsoft in our Secure Future Initiative update:
Latest Scott and Mark Learn To... podcast is out and it's a timely topic. We do some vibe coding and discuss its current limitations: Scott & Mark Learn To… ShaderGlass and Vibe Coding
ChatGPT Deep Research does indeed hallucinate when given a false premise (or when asked to do research on the ungrounded facts ChatGPT hallucinates). This is what I call "induced hallucination":
This was my first “internals” book. It includes the full ROM disassembly of Applesoft BASIC, which I used to figure out how to extend the language with a computed GOSUB command:
www.atarimagazines.com/compute/issu...
Reposted by
Last week, after nearly 17 years on the job, I resigned from the FBI.
I saw the threat Donald Trump and Kash Patel are to American safety and security and I couldn’t stand idly by any more. See my full resignation letter below.
open.substack.com/pub/fbisulli...
Reposted by
🦀 Hello World!
The Rust project now has an official presence on Bluesky! ✨
We'll be posting the same on our Mastodon and Bluesky accounts, so you won't miss anything on either platform.
Had dinnner with my good friend We decided LLM interfaces should be called HLIs (human language interface). More concise and precise than natural language interface.
Reposted by
This podcast is so good, I highly recommend it.
Love listening to and 👍
youtu.be/gUOExMkZCrM?...
Latest podcast episode: Scott and Mark Learn To...YAP!
Reposted by
Replied to
Here is one for you: 50's/60's space atomic age ads.
ALT